Unifying Graph-Based and Pairwise-Based Representations for Gene Regulatory Network Inference from scRNA-seq Data

Hang Li, Jiaxin Qi, Yan Cui, Jinliou, Yuhua Zheng, Jianqiang Huang

Keywords: AI for science

Abstract: Gene regulatory networks (GRNs) capture the underlying interactions through which transcription factors (TFs) regulate genes. Based on gene expression data, existing GRN inference approaches generally fall into two categories: graph-based methods, which model the GRN as a whole graph, and pairwise-based methods, which decompose the GRN into individual TF–target gene pairs for modeling. However, each approach exhibits limitations that are precisely the strengths of its counterpart. Graph-based methods tend to overfit due to their reliance on a single training graph, compared to the numerous TF–target gene pairs available for pairwise-based methods during training. In contrast, pairwise-based methods overlook the global topological structure, which is essential to graph-based learning. To address these limitations, we propose scUniGP, a unified framework that jointly models global regulatory topology and local TF–target interactions. scUniGP first extracts multi-scale topological features from the whole regulatory graph, and then hierarchically integrates these global representations with local features derived from pairwise modeling for comprehensive GRN inference. Extensive experiments on seven benchmark datasets demonstrate that our model consistently achieves state-of-the-art performance, validating the effectiveness of our integrative design.

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Submission Number: 8402